EP2885725B1 - Updating cached database query results - Google Patents
Updating cached database query results Download PDFInfo
- Publication number
- EP2885725B1 EP2885725B1 EP13750848.7A EP13750848A EP2885725B1 EP 2885725 B1 EP2885725 B1 EP 2885725B1 EP 13750848 A EP13750848 A EP 13750848A EP 2885725 B1 EP2885725 B1 EP 2885725B1
- Authority
- EP
- European Patent Office
- Prior art keywords
- query results
- database query
- computed
- computation
- platform
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 claims description 34
- 238000004590 computer program Methods 0.000 claims description 2
- 101000844801 Lactiplantibacillus plantarum (strain ATCC BAA-793 / NCIMB 8826 / WCFS1) D-alanyl carrier protein 2 Proteins 0.000 description 34
- 230000004044 response Effects 0.000 description 20
- 230000010006 flight Effects 0.000 description 18
- 230000008859 change Effects 0.000 description 17
- 230000008569 process Effects 0.000 description 14
- 238000013459 approach Methods 0.000 description 12
- 238000004891 communication Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000013179 statistical model Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 6
- 230000003247 decreasing effect Effects 0.000 description 6
- 238000005070 sampling Methods 0.000 description 6
- 230000006399 behavior Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000015556 catabolic process Effects 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000003642 hunger Nutrition 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000009257 reactivity Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000037351 starvation Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
- G06F16/24534—Query rewriting; Transformation
- G06F16/24539—Query rewriting; Transformation using cached or materialised query results
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/02—Reservations, e.g. for tickets, services or events
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/14—Travel agencies
Definitions
- the present invention is directed to the field of database technology. More specifically, it concerns pre-computing and caching database query results and strategies of keeping these results up-to-date.
- a common problem in database technology is to ensure short response times to database queries which require processing large volumes of data. For example, such computing-power consuming processing has to be performed in response to so-called "open queries" which contain only little input information (e.g. only one or two parameters out of a dozen possible parameters are specified and/or the specified value ranges of the parameters are broad) and, consequently, lead to a large number of results in general. Possibilities to speed up data processing by increasing hardware performance are limited. Thus, attention is drawn to improving the mechanisms underlying the processing of large data volumes.
- WO 01/33472 concerns an availability system used in a travel planning system.
- the system includes a cache having entries of availability information regarding airline seats.
- a cache manager manages entry information in the cache in order to keep information in the cache correct, current, complete or otherwise as useful as possible.
- the cache manager determines if a stored answer is stale and, if this is the case, sends an availability query to a source of availability information.
- Cache entries to be modified are obtained by asynchronous notifications from external systems and determined by a deterministic, predictive or statistical model.
- WO 02/25557 pertains to an information retrieval system wherein information received from information sources is cached for future use, such as for future client requests.
- Proactive queries can be generated to populate a cache and/or to update presently cached information.
- proactive queries are ordered on the basis of statistics or predictive indications such a nearness of departure time, the age of cached data, remaining seats in an aircraft, holidays or special events or equipment type.
- updates are received by external notifications from airlines such as AVS messages.
- EUI of an object is e.g. updated when the object itself is updated by the server or the object is not updated after its estimated mean refresh time
- US 6,725,333 relates to managing entities such as an object, an image file, or a webpage, stored in a cache and/or entities which may be stored in a cache.
- Program code is analyzed to determine if there is at least one statement which affects a desirability of performing at least one cache transaction and, if it is desired the at least one cache transaction is performed.
- a program analysis tool statically analyzes program code to locate points where object state changes occur, where objects are created and where objects are deleted, and then generates regularized dependencies at such points for and employing the dependencies to invalidate dependent cached queries. More specifically, the program is analyzed to identify one or more statements which may modify a value of one or more cachable entities during run-time of the program.
- a probability is determined which represents the likelihood that the detected statements will be executed, i.e., the likelihood that one or more cachable entities will change due to execution of the statements.
- a cache transaction e.g., inserting an object in cache or deleting or updating a cached object
- a determination is made as to whether the probability of change of one or more cachable entities meets a predefined threshold. If so, the system may be in favor of not caching one or more uncached entities and/or be in favor of invalidating or updating one or more cached entities. If not, the system may be in favor of caching one or more uncached entities and/or not be in favor of invalidating or updating one or more cached entities.
- a method of updating pre-computer database query results in a distributed database system is provided as defined in the independent claim 1.
- a database cache platform is provided as defined in the independent claim 13.
- a non-transitory computer readable storage medium is provided as defined in the independent claim 15.
- database query results corresponding to expected queries are generally pre-computed and cached (subsequently, the term "query” is used as a general term including any type of information retrieval requests such as transactional queries, requests for batch computations and other forms).
- the cached results are stored and are returned to the querying entity in response to actually occurring queries.
- Fig. 1 illustrates such a database system 1 on an abstract level.
- Basic data is kept in a computation platform 3 which is connected to a cache platform 2.
- the latter issues re-computation orders to the computation platform 3 which, in turn, transmits the corresponding results back to the cache platform 2 where the pre-computed query results are stored.
- Crafting re-computation schedules to determine which query results are to be re-computed at which times manually, by a human administrator, may prove efficient for specific purposes, but it is rigid and inflexible.
- the schedule needs to be re-crafted again when the assumptions and conditions underlying the schedule change. It also cannot dynamically track a sudden degradation of the cache quality which could arise in the event of massive changes in the underlying data domain. It is also hard to design such a schedule manually, e.g. due to missing objective quality criteria, and to maintain in terms of human costs.
- a still further approach is to re-compute data when they are getting too old. Depending on the nature of the underlying data and the query results to be pre-computed, it may, however, be difficult to evaluate good thresholds of "oldness".
- the strategy of updating the cache presented herein relies, as a first aspect, on means to estimate the accuracy of the entire cache of pre-computed database query results on the basis of a predictive model. As a second aspect, it is also checked whether those estimations are generally in line with reality, verifying that the model-based re-computation strategy is still valid on the occurrence of actual real-time (and real-life) events which may serve as indications that, for example, a significant part of the data underlying the cached queries has been changed and - due to these changes - also the corresponding cached queries are out-of-date.
- the predictive model models the discrepancies between the cached query results and presumed actual database query results, i.e. it approximates the accuracy or inaccuracy of any cached query result.
- the model models, for example, the probable volatility of the cached results over time. Presumptions on the cached results' volatility are concluded and extrapolated from (past) real-world experiences on the subject-matter of the respective data domain.
- the predictive model is generally a representation of the reality in terms of volatility and/or accuracy of the pre-computed database query results.
- the predictive model may be a conceptual model which has been generated by a structured modelling process known as scientific modelling. Modelling requires identifying and selecting those aspects and events of the real-world environment (here e.g.: the aspects and events having influence on the volatility and accuracy decreasing over time of the pre-computed and cached query results) and leaving aside other aspects which may not particularly relevant for the model's purpose.
- the identified and selected aspects may be abstracted and formulated/implemented in processible form such as a computer program and/or mathematical equations.
- Such a conceptual model may also serve as a means for simulation simulating the behavior (i.e. the volatility / accuracy) of the pre-computed database query results over time.
- the employed predictive model may be a historic-statistical model which is based on monitoring and evaluating e.g. the volatility / accuracy of the pre-computed query results in the given period of time in the past (e.g. three months) and extrapolates the determined trends into the future.
- a historic-statistical model is given further below.
- the underlying data may be located in the domain of air travel and contain information on flights such as departure and destination airport, airline, departure and return dates, fares, booking classes and the like.
- This air-travel related data is kept in the computation platform and is queried by customers in order to get knowledge of availability and prices of air flights.
- Computing prices on the basis of the basic flight data is resource- and time-consuming.
- the actual prices are pre-computed and cached in the present cache platform.
- the probabilistic model models the volatility of the flight prices over time.
- the required knowledge to build such a model can be taken from real-world experiences on the behavior and development of flight prices prior to the departure date. For example, it might be known that flight prices remain relatively stable over the time period prior to one month before the respective departure dates, but get more volatile during the month before the departure date. Hence, the probabilistic model indicates that pre-computed cached prices belonging to flights upcoming in the next month should be re-computed more often than such pre-computed prices which are associated with flights in the more distant future.
- the cache data platform 2 is equipped with external interfaces to communication sources which notify the cache data platform accordingly with the relevant information.
- Such real-time events may pertain to particular situations which are not considered in the predictive model. For example, a portion of the cached prices may be affected by a promotion, whereas other prices may go more volatile at a particular time of the year (such as holiday seasons, Christmas etc.). Also "exceptional" situations like a trade fair, a sport event or the like, random events such as strikes or natural disasters could change the presumptions underlying the "normal" model causalities. These particular influences can be considered when determining the probabilities that cached query results are outdated in response to respective real-time events representing such exceptional situations. Alternatively, the impact of scheduled events such as trade fair, holiday seasons, sports events and the like can be introduced into the probabilistic model in good time before the date of the event.
- the update strategy presented herein is capable of taking into account "uncertain" events, i.e. such events which do not invalidate one or more pre-computed cached query results with certainty, but only indicate that the probability might be increased that cached database query results are outdated.
- these events are indeterministic with regard to the accuracy of cached query results and only have a probabilistic influence on the discrepancies between cached query results maintained in the cache platform 2 and presumed actual database query results resulting from a hypothetical re-computation.
- This is different from the proposals described in WO 01/33472 and WO 02/25557 wherein the AVS messages, for example, indicate that a particular flight has been cancelled. Consequently, on reception of such an AVS message, it is known with certainty that the respective airplane seats are not available anymore.
- a real-time event having only a potential impact on the accuracy of the cached query results could be a fare update.
- a fare is a set of data including parameters like departure and destination city, booking class, flight type (one-way or round-trip), an amount of money and rules which define constraints that must be met for the fare to actually apply.
- fares represent the basic data for price calculation of a particular flight. If fares for a specific origin-destination-city-pair are updated by the airline, the likelihood may be increased that a precalculated and cached flight price regarding this city pair is not correct anymore.
- the fare which was actually applied by the pre-computation platform 3 when pre-computing the cached price is unknown to the data cache platform 2.
- the fare applied for the previous pre-computation might have actually not been changed and the fare changes indicated by the fare change event do not change the fact that the previously pertinent fare still applies and, therefore, the price calculated before remains valid.
- the previously applied fare is actually changed, but - due to the change - another fare now applies for calculation of the flight price in question which, in the end, has the effect that the cached price actually remains valid.
- the data cache platform 2 can only guess with an indeterministic likelihood that certain cached query results are now outdated and it would be advantageous to re-compute them in order to keep the cache accurate.
- this is not a certain fact and it may well be that the respective cached query results - although their probability of being outdated has increased - are actually still accurate.
- the determination of probabilities that cached database query results are outdated is performed in two logical steps: Generally, at the first logical level, the probabilities are identified by using the probabilistic predictive model. Subsequently, at the second logical level, these determined probabilities may be amended in response to incoming real-time events.
- the data cache platform 2 automatically generates and issues re-computation orders to the re-computation platform 3 via an appropriate network interface between the two entities (cf. Fig. 1 ).
- re-computation orders are generated with respect to those cached query results which have a higher probability of being outdated than other cached query results which have a lower probability of being outdated.
- This general rule of thumb may be implemented by using threshold values of the probabilities: Cached query results with a determined probability of being outdated above such a threshold value, need to be re-computed. Accordingly, respective re-computation orders are sent out.
- Cached query results with a determined probability of being outdated at or below such a threshold value are considered as likely being still accurate and consequently do not need to be re-computed. Accordingly, no re-computation orders are issued with respect to these cached query results.
- the available computation capacities at particular times are taken into account by the data cache platform 2 before sending out the re-computation orders.
- the data cache platform 2 needs to have knowledge about the degree and/or schedule of the computation platform's 3 capacity utilization and free computation resources, respectively.
- the relevant information are populated via the communication link between the two platforms.
- the re-computation platform 3 In response to receiving a re-computation order, the re-computation platform 3 re-computes the respective query results and returns them to the data cache platform 2 where they are stored and the monitoring and determination of the probabilities recurs.
- This determination whether or not and to which extent an event is present in the probabilistic model generally depends on the type of the model used. If, for example, the model is a conceptual model being the result of a scientific model generation process (as briefly outlined above), the skilled person having generated this model is aware which aspects, events and assumptions have been considered to be relevant for and incorporated into the model and which other aspects, events and assumptions have been regarded to be of less relevance and have therefore not been included in the model. In this case, it is therefore generally known which asynchronous real-time events are represented in the model and which are not.
- the model is a historic-statistical one which is mainly based on past developments and trends of the volatility, accuracy and/or popularity of the pre-computed database query results, it may not known from the outset which events have led to this particular behavior of the pre-computed query results' volatility / accuracy /popularity in the past.
- One approach is, for example, to determine the actual accuracy of pre-computed database query events (e.g. by re-computing a representative sample of them and comparing the re-computation results with the cached query results) and compare it with the predicted accuracy indicated by the probabilistic model.
- the probabilistic model is unaware of the particular asynchronous real-time event, i.e. the event is not represented in the model (otherwise, the model would also have indicated a lower accuracy). If, on the other hand, the actual (sample) accuracy is in the area of (i.e. only insignificantly below, at or even higher than) the accuracy predicted by the probabilistic model, it can be assumed that the asynchronous real-time event has been included in the probabilistic model (or it is only a negligible event having only limited influence on the volatility or accuracy of the pre-computed database query results). A more detailed example of this approach is given further below.
- occurring real-time events which are present in the probabilistic model to some extent are accumulated in order to assess trends. If an accumulation of actually emerging real-time events generally modeled by the probabilistic model indicates a burst the extent of which is beyond what is considered by the model probabilities are amended and, if applicable, re-computation orders are overridden accordingly.
- events are also accumulated and analyzed in groups in order to filter out events which might outdate too few cached query results and/or might be considered irrelevant. Also for this reason, events are logged, collected over time and handled in an aggregated manner. In this way, generating too much re-computation orders in response to low-impact events is prevented and, thus, a disproportional increase of computation resource costs is avoided.
- the present cache update strategy can, for example, be used in conjunction with Amadeus' Massive Computation Platform (MCP) which is subject EP 2521074 A1 .
- MCP Massive Computation Platform
- Database query results such as travel recommendations generated by the MCP are duplicated and stored in the data cache platform for further analysis.
- Re-computation decisions are made on the basis of a probabilistic model which itself may be composed on the basis of statistics data taken from other Amadeus services.
- real-time events such as flight fares changes, airplane seat availability changes, booking class invalidations, client flight ticket requests, user quality feedback events, flight cancellations and/or the like are taken into account.
- pre-shopping Before actually booking a travel, end users of the travel industry normally want to inform themselves about available flights including current flight prices, without any commitment to actually book the flight. Very often, such non-binding requests for pre-booking information take the form of open and broad database queries which would require an enormous amount of computing resources when computed only at the time of query. Moreover, customers expect the requested information to be delivered virtually instantaneously in response to their queries. Thus, pre-shopping query results such as priced air travel recommendations are usually pre-computed and cached. Consequently, travel industry pre-shopping forms a suitable application for the cache update strategy proposed herein.
- Fig. 2 shows an overview of the distributed database system 1 according to an exemplary embodiment.
- the embodiments described subsequently relate to databases in the travel industry. Specifically, an embodiment is presented in which the computation platform 3 maintains data on air travel offers and the cache data platform 2 stores prices related to these air travel offers which the computation platform 3 calculates on the basis of calculation rules, in particular flight fares and their associated calculation rules.
- these embodiment are examples only for the purpose of illustrating the present cache update strategy in more detail.
- the cache update strategy presented herein can be applied to any kind of data and database query results independent from the structure and/or semantics of the data and the cached results.
- the main entities of the distributed database system 1 are the data cache platform 2 (hereinafter briefly referred to as DCP) and the computation platform 3.
- the computation platform 3 is a Massive Computation Platform (MCP) as disclosed in the European Patent application 11305518 .
- MCP Massive Computation Platform
- the DCP 2 and the MCP 3 are coupled via at least one communication link which is/are utilized to transmit re-computation orders from the DCP 2 to the MCP 3 and, in response, pre-computed priced travel recommendations (hereinafter also briefly referred to as "prices") back from the MCP 3 to the DCP 2.
- the DCP 2 is equipped with additional communication interfaces for incorporating data which it uses for the determination of the accuracy probabilities of the cached prices.
- These interfaces include, for example, communication links for incorporating statistical data which forms the basis of the probabilistic model and for receiving asynchronous real-time events such as fares changes and flight availability announcements populated by airlines or customer promotion campaigns.
- the distributed database system 1 may comprise application platforms 4 that organize and maintain data which can be queried by end users or external customers such as travel agencies.
- Application platforms 4 are populated and updated by the MCP 3 via respective communication links between MCP 3 and application platforms 4. This population and update is triggered by the re-computation orders issued by the DCP 2.
- the MCP 3 in response to the re-computation orders received from the DCP 2, the MCP 3 re-calculates the prices of the travel recommendations and returns them to the DCP 2. Simultaneously, however, the MCP 3 also forwards the recomputed priced travel recommendations to the application platforms 4 and which store them as well (as indicated by "travel recommendation integration" in Fig. 2 ). Consequently, also the application platforms 4 cache the pre-computed priced travel recommendations which are queried by users, on the basis of the cache update strategy implemented by the DCP 2. Thus, the present cache update strategy is leveraged to applications which provide benefits to users, e.g. in form of instant responses to open queries.
- the data cache platform 2 acts as a control platform arranged to control and trigger the updates of the application platforms' 4 caches.
- the cached database query results stored in the data cache platform 2 are not actually accessed or queried by any users or customers, but merely form the control data basis on which the cache update strategy is performed.
- the data cache platform 2 may also be directly queried by users or customers, or - in other words - the present cache update strategy may be implemented directly in one or several application platform(s) 4 as opposed to a separate control entity.
- the application platforms 4 comprise, for example, a pre-shopping application platform, a fares analysis application platform and other platforms.
- the pre-shopping application platform is queried by end users who wish information about flight availability and prices. For example, an end user could direct a query to the pre-shopping application in order to obtain an overview of prices of travel offers during the holiday season departing from Nice below 500 Euros. Due to the pre-computed priced travel recommendations cached in the pre-shopping application which are updated in line with the present cache update strategy, prices of the respective flights do not need to be computed at the time of the occurrence of the query. Rather, a list of travel offers fulfilling these rather unspecified constraints can be returned very fast on the basis of the priced travel recommendations cached in the pre-shopping application platform.
- the user can then select a travel from the returned list that suits him/her best and can then issue a further request for actually booking this travel.
- the second request is then processed by a booking engine (not shown) which calculates the current and actual price and suggests a binding offer to the user.
- the data cache platform 2 is composed of three modules:
- the DCP 2 includes an internal database 8 which keeps the cached priced travel recommendation data.
- This representation only retains attributes of the price information which are relevant for assessing the out-of-date probabilities and deciding on the re-computation decision, such as, for example: city pair, the departure date, stay duration, and date of last computation, all of which being outputs of computations returned by the MCP 3.
- Other data utilized by the MCP 3 in order to perform its computations such as fares are not mirrored to the DCP 2 as they are not necessary to perform the cache update strategy.
- the DCP 2 enriches its data with meta-data attributes (which are not part of the data sets returned by the MCP 3), such as an initial presumed accuracy (chances that a price just re-computed by the MCP 3 differs from calculation for booking), volatility (indication of the probability that a price differs from calculation for booking since its last computation) and a popularity (how often the flight is searched for and booked).
- meta-data attributes represent the probabilistic model (and therefore depend on the specificities of the probabilistic model) on the basis of which the analyzer 6 determines the probabilities whether cached prices might be outdated (as will be explained in more detail below).
- the input manager 5 is arranged to convert and integrate all the heterogeneous sources of information into the local representation database 8 of the prices returned by the MCP 3. It records events and actions that potentially impact modeled prices. These events include customer promotions and customer discrepancies feedback. Furthermore, flight availability events like flight cancellations do not only invalidate the cached travel recommendations which are directly based on the canceled flight, but may also influence parallel cached data such as flights of the same city pair which are scheduled at the same time of the cancelled flight. These real-time events are then forwarded to the consolidator 7 which processes them further in order to amend out-of-date probabilities and re-computation decisions.
- the priced data cached in the DCP 2 is globally mapped to the underlying data domain kept in the MCP 3 while reducing costs for storage resources significantly.
- Probabilistic encoding is, for example, implemented by using Bloom Filters. The effect of such encoding is twofold. First, Bloom filters are conservative. They allow to positively track at least and in any case those prices which are probably impacted e.g. by a real-time event indicating fares changes, but they are not wrong on the reverse, prices which are considered to be not affected are actually not affected. So there is no risk of failing to recognize prices potentially influenced by such a fares change event. Secondly, the amount of false positive indications strictly depends on the allocated size of the Bloom Filter, so one can limit their occurrence as needed.
- the second module performs the first, general level of determination of the probabilities whether the cached priced travel recommendations are outdated based on the model of probabilistic degradation of the accuracy of the pre-computed prices kept in the DCP 2. It examines and evaluates the meta-data added to the prices by the input manager 5, as explained above.
- the probabilistic model represented by this meta-data thus includes metrics regarding price volatility included by the volatility database 10, initial accuracy of prices incorporated from the initial accuracy database 11 and the popularity of flight recommendations as returned by popularity reports from the statistics servers 9. It outputs probability and priority information regarding the cached prices to the consolidator 7, i.e. indications which prices need to be recomputed with priority based on the static probabilistic information only (i.e. without considering any event).
- an exemplary probabilistic model is based on the following parameters for a pre-computed database query result x i :
- These parameters of the probabilistic accuracy model are for example set according to statistical history data derived from the statistics server 9 based on several days, weeks or months of history. Every single pre-computed database query result is modeled with these parameters in order to predict the state of the pre-computed database query results and thus to predict the quality of the whole cache.
- the analyzer 6 is configurable to apply those strategy or a strategy mix depending on the situation (e.g. in accordance to an agreement with the customer owning the underlying travel data in the MCP 3, depending on the amount of data, depending on the available computation resources, depending on the objection in which way the cache should be optimized).
- the following strategies may be applied:
- the analyzer 6 updates the volatility model database 10 based on the values of recently re-computed prices received from the MCP 3 and incorporated into the DCP 2. As the analyzer can track the actual volatility of cached prices based on repeating re-computations, it can feed these statistical information back to the volatility model database 10. To update the volatility model, the analyzer 6 counts the number of differences between the newly computed price results and the previously received price values. From these differences it updates the volatility parameters for the respective parts of the analyzed prices.
- the analyzer 6 may update the initial accuracy database 11 in the same manner. It may also ask for other popularity reports, for example, if prices from new city pairs have been integrated in the DCP 2 for the first time.
- the analyzer 6 performs its processing with default parameters to be as conservative as possible.
- the consolidator 7 performs the second level of the probability determination by taking into account incoming real-time events.
- it is the entity which generates the re-computation orders and issues them to the MCP 3. It takes the outputs of the analyzer 6 as a base for its decision. These provide a first estimate of the re-computation priorities for all prices of the data domain. Then, it overlays all the information gathered from the various sources of real-time events to amend the re-computation priorities. This results in enhanced re-computation priorities.
- the consolidator may consider any customer service level agreement such as, for example, "guarantee that all prices will be recomputed at least once every week", and amend the priorities accordingly. It selects those entries in the internal prices data representation 8 with the highest priorities first and marks them for re-computation. As the consolidator preferably has knowledge of the computation resources available at the MCP 3, it is able to earmark as many cached prices as can be re-computed by the MCP 3 in a particular time interval. It then sends the resulting re-computation order to the MCP 3.
- customer service level agreement such as, for example, "guarantee that all prices will be recomputed at least once every week”
- Information from real-time events is a means to improve the accuracy of the cached data over strict statistical modelling. They can be used to track what is really happening instead of only what was expected. It is a means to control the predictions of the statistical model and amend them/it when the predictions are proved wrong or inappropriate.
- Actors' events generally occur selective (i.e. from time to time), but may have a drastic influence on the re-computation decisions. Externals customer may provide feedback on the discrepancies between cache and shopping that he is experiencing on his own platform. This feedback can be used to amend the accuracy predicted by the statistical model and thus force sooner re-computations when needed.
- a provider of data stored in the MCP 3 e.g.
- a travel provider offering travels performs a promotion campaign promoting flights on certain city pairs, it can be assumed that these prices are more volatile and get outdated more often. Thus, the re-computation frequency of these prices during the promotion campaign might be needed to be increased.
- a maintenance operation for the MCP 3 may be necessary from time to time or operational issues may be experienced within the system 1. In such cases, the DCP 2 can be commanded to generate less or no re-computation orders until the maintenance is done and the issue is recovered, respectively.
- Availability events indicate a real-time status on the accuracy of cached flights.
- other prices may be affected, wherein the effect is uncertain and, therefore, the probability that these prices are outdated may be increased.
- a "class closed" event indicates that a particular booking class has become full for a particular flight. Seats in this flight and class are no longer bookable and thus the respective prices cached by the DCP 2 have become invalid with certainty.
- this might be considered as an indication that other classes of the same flight and/or seats in the same class in other flights departing shortly before or after this flight might get more volatile.
- fares contain information and logic such as rules which are used to calculate a price of a particular flight.
- a set of fares is accessed, it is decided which fare is relevant and is actually to be applied and, finally, the price is calculated.
- fare(s) which, however, also may change over time, as the constraints which fare applies to a particular flight can change.
- relations in the other direction "fare ⁇ flights" are generally not tracked, i.e. it is not clear which fare does apply to which flights.
- a change in a fare potentially impacts a huge amount of prices calculated from the underlying data domain.
- the communication between the MCP 3 and the DCP 2 can be utilized to provide the DCP 2 with a mapping to fares applied by the MCP 3 for computing prices.
- the MCP 3 records all fares accessed for calculating the requested prices. This information is then stored globally in a mapping fares ⁇ flights and maintained during every computation by the MCP 3.
- the input manager 5 searches this global mapping in order to determine the flights impacted by the fares change event and marks them as "updated". Note that a fare change does not necessarily imply a price change for a travel recommendation, as briefly explained above.
- the consolidator 7 first assesses the potential influence of the real-time event on the cached prices rather than initiating re-computation of cached travel recommendations without having considered the relationship of the event with the basic probabilistic model. Such events are first analysed with respect to their representation in the probabilistic model.
- the specificities of this analysis depend on the type of probabilistic model used.
- incoming events are checked against the characteristics of the probabilistic model known from the model generation process in order to determine whether or not a particular event is present in the model.
- a historic-statistical model may be used and it may accordingly not be known a priori which events have led to the historic-statistical data basis of the model.
- the following approach may be taken in order to determine whether or not a particular event is represented in the probabilistic model (for completeness: The following approach is not limited to any particular type of probabilistic model and can also be employed e.g. for conceptual models).
- the cached database query results are subdivided into smaller portions, referred to as "shares" hereinafter.
- the shares are, for example, formed by origin-destination pairs, i.e. one share includes all pre-computed priced travel recommendations having the same origin and destination location.
- the data domain is only of limited size, no subdivision is performed. In other words, in this case, there is only one share covering all pre-computed database query results kept in the DCP 2.
- one share or more shares potentially affected by this event i.e. which contain pre-computed database query results which may have an increased probability of being outdated due to the detected event, are identified.
- An actual accuracy of the at least one share is determined, e.g. by estimating its accuracy by re-computing a particular or random sample of pre-computed database query results in the share. For example, if the share contains 10,000 pre-computed database query results, a sample of 20 pre-computed database query results is re-computed. This re-computation is performed similar to other re-computations.
- the DCP 2 generates and issues re-computation orders to the MCP 3.
- the MCP 3 performs the re-computations and returns their results, i.e. the updated sample pre-computed database query results to the DCP 2.
- the DCP 2 compares the results of the sample re-computations with the corresponding database query results previously cached and, e.g., determines the percentage of the sample pre-computed database query results having changed (i.e. which have been outdated or inaccurate) and/or the percentage of the sample pre-computed database query results being unchanged (i.e. which have not been outdated and were still accurate). These percentages of the sample pre-computed database query results might be transferred to the overall share as an estimation of the overall share's accuracy, i.e. it could generally be concluded from these sample percentages that the overall share is similar (in-)accurate as the sample.
- the determined actual accuracy of the share is compared with the share's accuracy predicted by the probabilistic model. Depending on the outcome of this comparison, it is concluded whether or not and, possibly, to which extent the particular asynchronous real-time event is represented in the probabilistic model. If the determined actual accuracy is below the accuracy predicted by the probabilistic model to more than a given extent, i.e. the sample accuracy is substantially lower than the accuracy predicted by the model, the event is not considered to be included in the probabilistic model. If, on the other hand, the determined actual accuracy is below the accuracy predicted by the probabilistic model less than the given extent, or both accuracies are identical or the determined actual accuracy is even higher than the predicted accuracy, the event is considered to be represented in the probabilistic model. The difference between both accuracies is also an indicator of the extent to which the event has been incorporated in the model.
- sampling itself can be recognized as a way to detect occurred asynchronous real-time events in the first place.
- sampling may be utilized to implicitly detect asynchronous real-time events which are not explicitly indicated by a third party.
- a continuous or regular sample process over all shares may employed in order to estimate the actual accuracy of all shares of pre-computed database query results. This allows implicitly detecting events not reflected in the probabilistic model if the comparison of the determined actual accuracy and the accuracy predicted by the probabilistic model turns out that the first one is substantially lower than the latter one.
- the identification of one share or more shares potentially affected by a detected event depends on the characteristics of and knowledge about the events. If, for example, an event has been implicitly detected by a cyclic sampling process, the potentially affected shares have been determined at the same time (namely those shares which determined actual accuracy is substantially lower than the predicted accuracy).
- an explicit signaling of an event may also contain an indication of potentially affected pre-computed database query results. If, for example, an airline has signaled cancellation of a particular flight, the prices of flights between the same origin and destination during times before and after the cancelled flights might be considered to be potentially affected.
- the actual accuracy of one pre-computed database query result x i is referred to as Acc i .
- the collection of all pre-computed database query results can be logically split into shares D j .
- the following considerations relate to one specific exemplary share D .
- the average accuracy of the pre-computed database query results contained in share D as predicted by the probabilistic model is denoted as Acc model (D) .
- the proportion of actually accurate pre-computed database query results in D is referred to as Acc(D) .
- AC N The number of accurate results among the N is referred to as AC N , i.e. after having re-computed the N sample pre-computed database query results, AC N re-computed database query results are identical with their previously cached correspondents while (N - AC N ) re-computed database query results differ from their previously cached correspondents and have therefore been found inaccurate.
- the accuracy probability for one pre-computed database query results x i within share D is Acc(D) .
- the relation A C N N is an estimation of the actual accuracy of share D , Acc(D) , which is conveyed by the result of the sample re-computation of the N sample pre-computed database query results.
- Acc model (D) is the accuracy estimation of Acc(D) given by the probabilistic model.
- H 0 The prediction of the probabilistic model appropriately indicates the accuracy of share D , i.e. Acc(D) ⁇ Acc model (D) .
- a C N N is lower than Acc model (D) due to statistical deviations caused by the sampling process (i.e. accidentally, the selected N pre-computed database query results are more inaccurate that the average of all pre-computed database query results in D and therefore the N samples do not representatively reflect Acc(D)).
- Taking the decision D 0 that H 0 is true or the decision D 1 that H 1 is true can be done the following way: If the actual accuracy is lower than the predicted accuracy to more than a given extent, there is an increased likelihood that the accuracy predicted by the probabilistic model is inappropriate and needs to be reduced. This is, for example, the case if Acc model (D) indicates an accuracy Acc(D) of 85%, whereas the actual accuracy received by the sample re-computation only indicates only 20% of accurate pre-computed database query results within D . In this case, D 0 is the right decision. If, however, the actual accuracy received by the sampling and the accuracy predicted by the model are similar, i.e. the deviation between both indications is relatively small, the accuracy predicted by the probabilistic model is probably appropriate. To give an example, this may be the case if Acc model (D) indicates an accuracy Acc(D) of 85% and A C N N amounts to 80% accuracy. In this case, D 1 is the right decision.
- a threshold K Th is introduced: If AC N > K Th , then the accuracy predicted by the probabilistic model Acc model (D) is considered as an appropriate approximation of Acc(D) , D 0 is chosen. If AC N ⁇ K Th , then the accuracy predicted by the probabilistic model Acc model (D) is not considered to be an appropriate approximation of Acc(D) , D 1 is chosen.
- K Th it is advantageous to have information about the quality of the probabilistic model.
- K Th One exemplary way to set K Th by taking into account the past quality / reliability of the probabilistic model is to consider the probability of an erroneous decision D 1 , i.e. Acc model (D) is not regarded to be an appropriate estimate of Acc(D) although it is actually one.
- H 0 K N ⁇ Ac c model D K ⁇ 1 ⁇ Ac c model D N ⁇ K
- H 0 P A C N ⁇ K Th
- H 0 ⁇ K ⁇ K Th K N ⁇ Ac c model D K ⁇ 1 ⁇ Ac c model D N ⁇ K
- This formula allows to choose ⁇ and define K Th in accordance with the past reliability of the probabilistic model. If, for example, the probabilistic model has been correct for 99% of the time, ⁇ is set to 0,01. The threshold value of K Th can then be calculated using the previous formula. If the number of K accurate sample re-computation results is lower than K Th , i.e.
- the events are processed as soon as possible.
- this amendment is optionally achieved by applying the estimation for Acc(D) gained by the sample re-computations, i.e. A C N N , to the probabilities of the pre-computed database query results in the share D being outdated.
- the decision which pre-computed database query results are to be re-computed is taken as usual, i.e. in the same way as if no amendments of the probabilities had happened.
- the probability stated by the probabilistic model for the accuracy / outdateness is e - ⁇ i t i .
- it can be amended in response to recognized asynchronous real-time events e.g.
- the next accuracy computation by the analyzer 6 starts again from the previous value (which has potentially been amended in response to real-time events).
- the amendment of the probabilities in response to incoming asynchronous real-time events e.g. the accuracy predicted by the probabilistic model, is reset after the respective pre-computed database query results are re-computed by the MCP 3 the next time.
- Events which are represented in the probabilistic model at least to a certain extent like fares or availability changes, are optionally accumulated and their appearance is regularly compared with the predictions of the probabilistic model.
- event peaks do not match the model locally, i.e. a burst of events is significantly outside the statistic underlying the probabilistic model, the impacted prices are marked as potentially outdated in order to re-compute them as soon as possible.
- "noise” caused by events already present in the probabilistic model and therefore already considered by the determinations conducted by the analyzer 6 beforehand is filtered out.
- the handling of incoming asynchronous real-time events is done by a dedicated component functionally separate from or a subcomponent of the consolidator 7, an event manager module.
- the events are received by the input manager 5 and initially stored without aggregation in the internal data repository.
- the event manager receives any incoming asynchronous real-time event, processes the event, e.g. by applying the strategy described in detail above, and possibly performs the amendment of the probabilities that the affected pre-computed database query results are outdated / inaccurate.
- the strategy described above is only applied if a given number of events has been recognized by the event manager applies the strategy described above and amend the probable accuracy of each pre-computed database query results is amended accordingly only then, if necessary.
- the consolidator 7 works on a grid-view of the internal data representation 8, i.e. it considers groups of adjacent prices together during its algorithm instead of working on the set of prices in an isolated manner.
- adjacent price data sets are seen as a single data set with aggregated properties' values.
- Working on aggregated data sets limits the generation of sparse re-computation orders, and thus increases mutualization and optimization opportunities at the MCP 3. This reduces computation costs.
- Fig. 5 summarizes the above detailed description and gives an overview of the cache update method presented herein.
- the process of keeping the cache of pre-computed database query results up-to-date starts with the determination 14 of probabilities of inaccurateness of the cached query results.
- This determination is composed of two activities which are located on two logical levels: Firstly and generally, a predictive model based on statistical and probabilistic data is employed in order to estimate the likelihood that a particular cached query result does not correspond to the query result (hypothetically) re-calculated. Secondly and more specifically, real-time events are taken into account which potentially impact and increase, respectively, the probability that cached query results are outdated. These real-time events are characterized by the fact that they, in general, do not indicate an inaccuracy of particular cached query results with certainty, but are indeterministic in this respect. Thus, on their occurrence, one can only draw probabilistic conclusions on the likelihood of accuracy and inaccuracy, respectively.
- the database re-computation orders are automatically issued 15 by the DCP 2. These orders are received by the MCP 3 which then re-computes the respective results and returns them 16 to the DCP 2. In turn, the DCP 2 receives the results and stores them 17 in a local representation 8. This concludes one update cycle and the next cycle re-iterates with the probability determination 14.
- the DCP 2 is configured to generate re-computation orders every 20 minutes, i.e. a round or cycle of determining probabilities whether cached data is out-of-date and generating and issuing re-computation orders takes 20 minutes.
- the resources at the MCP 3 are known a priori for a whole day and the DCP 2 is aware of the computation resources available at the MCP 3 and is therefore able to synchronize the amount of re-computations with the MCP's 3 available resources.
- the DCP 2 analyzes the current accuracy of the cached database query results, i.e. the priced travel recommendations stored in the internal database 8.
- the round will yield a set of re-computation orders that will be processed by the MCP 3 at the end of the 20-minutes round.
- orders from the last cycle are being computed and new price recommendation are generated and transmitted back to the DCP, where they are stored and available for analysis and update of recurring information in the next cycle.
- Fig. 6 shows that the MCP has significant available resources in the time interval between 04:00 and 05:00 a.m., so a lot of re-computations can be performed in this hour. Afterwards, however, no resources are available until 9:00 a.m., so no re-computation is possible at that time. Later during the day, from 09:00 a.m. to 7:00 p.m., some resources are available at the MCP 3.
- the analyzer 6 analyzes the cache accuracy, while the consolidator 7 generates re-computation orders accordingly. Those orders will be implemented by MCP 3 at 04:40 a.m..
- the analyzer 6 focuses the MCP price recommendation it received at the beginning of the round. It counts the differences between the received prices and the previous prices whose values have been stored in an internal repository. Based on the differences, it amends the "volatility" recurring source of information.
- the input manager 5 saves the received MCP prices for further inspection.
- the MCP 3 processes the re-computation orders received from the DCP 2 during the interval 04:20 to 4:40 a.m..
- the DCP 2 is aware that it cannot generate any re-computation orders for the time slice to come (05:00 a.m.) and its successors until 09:00 a.m. However, it still analyzes the data model continuously to update the current accuracy of all cache priced travel recommendations. It will do the same for every future cycle until 08:40 a.m.
- the analyzer 6 determines that the accuracy of the cache decreased during the previous four hours without any re-computation. It generates re-computation orders accordingly over the following cycles, but only to a less extent due to the limited amount of available resources at the MCP 3 from 09:00 a.m. to 7:00 p.m. Then, at 09:00 a.m., MCP 3 starts processing the new re-computation orders it received in the interval before (i.e. 08:40 to 09:00 a.m.) and will stop after the end of the round lasting from 6:40 to 7:00 p.m.
- the DCP 2 will not generate any further re-computation orders, but continue to analyze the cache accuracy on the basis of the probabilistic model and possibly incoming real-time events.
- Fig. 7 is a diagrammatic representation of a computer system which provides the functionality of the cache platform 2 of Fig.2 .
- the cache platform 2 includes a processor 101, a main memory 102 and a network interface device 103, which communicate with each other via a bus 104.
- it may further include a static memory 105 and a disk-drive unit 106.
- a video display 107, an alpha-numeric input device 108 and a cursor control device 109 may form a distribution list navigator user interface.
- the network interface device 103 connects the data cache platform 2 to the computation platform 3, the sources of statistical data needed to fill up the predictive model such as the statistics servers 9, the volatility database 10 and the initial accuracy database 11, the sources of real-time events, the Internet and/or any other network.
- a machine-readable medium on which the software 110 resides may also be a non-volatile data carrier 111 (e.g. a non-removable magnetic hard disk or an optical or magnetic removable disk) which is part of disk drive unit 106.
- the software 110 may further be transmitted or received as a propagated signal 112 via the Internet through the network interface device 103.
- the present cache update strategy provides a means to automatically generate cache re-computation decisions. It determines which cached query results are to be re-computed and controls the re-computation also time-wise by taking into account the available computation resources at the computation platform. Thus, in general, the accuracy of the cached query results is estimated on the probabilistic model which models the up-to-dateness and out-of-dateness, respectively, over time. This out-of-date-analyses enable processing several billions of sets of data per hour underlying the re-computation of cached database query results.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Transfer Between Computers (AREA)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP13750848.7A EP2885725B1 (en) | 2012-08-14 | 2013-08-09 | Updating cached database query results |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP12368020.9A EP2698729B1 (en) | 2012-08-14 | 2012-08-14 | Updating cached database query results |
US13/585,286 US9235620B2 (en) | 2012-08-14 | 2012-08-14 | Updating cached database query results |
EP13750848.7A EP2885725B1 (en) | 2012-08-14 | 2013-08-09 | Updating cached database query results |
PCT/EP2013/002390 WO2014026753A1 (en) | 2012-08-14 | 2013-08-09 | Updating cached database query results |
Publications (2)
Publication Number | Publication Date |
---|---|
EP2885725A1 EP2885725A1 (en) | 2015-06-24 |
EP2885725B1 true EP2885725B1 (en) | 2018-12-12 |
Family
ID=49003740
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP13750848.7A Active EP2885725B1 (en) | 2012-08-14 | 2013-08-09 | Updating cached database query results |
Country Status (11)
Country | Link |
---|---|
EP (1) | EP2885725B1 (pt) |
JP (1) | JP6162240B2 (pt) |
KR (1) | KR101972199B1 (pt) |
CN (1) | CN104471573B (pt) |
AU (1) | AU2013304366B2 (pt) |
BR (1) | BR112014031390A2 (pt) |
CA (1) | CA2875735C (pt) |
ES (1) | ES2714676T3 (pt) |
IN (1) | IN2014DN10461A (pt) |
SG (1) | SG11201408814XA (pt) |
WO (1) | WO2014026753A1 (pt) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022035481A1 (en) * | 2020-08-10 | 2022-02-17 | Google Llc | Anticipatory pre-execution of data queries |
US11341137B1 (en) | 2020-12-04 | 2022-05-24 | Amadeus S.A.S. | Processing search requests |
EP4009188A1 (en) | 2020-12-04 | 2022-06-08 | Amadeus S.A.S. | Processing search requests |
WO2022117787A1 (en) | 2020-12-04 | 2022-06-09 | Amadeus S.A.S. | Processing search requests |
US11748348B2 (en) | 2019-02-14 | 2023-09-05 | Amadeus S.A.S. | Processing complex database querys |
US11829345B2 (en) | 2019-03-04 | 2023-11-28 | Amadeus S.A.S. | Distributed data processing |
US11947553B2 (en) | 2019-03-04 | 2024-04-02 | Amadeus S.A.S. | Distributed data processing |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
ES2689305T3 (es) * | 2014-02-13 | 2018-11-13 | Amadeus S.A.S. | Aumentar la validez del resultado de búsqueda |
WO2015120968A1 (en) * | 2014-02-13 | 2015-08-20 | Amadeus S.A.S. | Increasing search result validity |
US9984165B2 (en) | 2014-02-13 | 2018-05-29 | Amadeus S.A.S. | Increasing search result validity |
US9582536B2 (en) | 2014-02-19 | 2017-02-28 | Amadeus S.A.S. | Long-term validity of pre-computed request results |
US10395294B2 (en) | 2014-11-03 | 2019-08-27 | Amadeus S.A.S. | Managing pre-computed search results |
CN107004026B (zh) * | 2014-11-03 | 2020-09-22 | 艾玛迪斯简易股份公司 | 管理预先计算的搜索结果 |
EP3016000B1 (en) | 2014-11-03 | 2024-07-31 | Amadeus S.A.S. | Managing pre-computed search results |
EP3128441B1 (en) | 2015-08-03 | 2018-10-10 | Amadeus S.A.S. | Handling data requests |
US20180075391A1 (en) * | 2016-09-15 | 2018-03-15 | Travelator Inc. | Network-based real-time enterprise travel management apparatus, methods, and systems |
CA3038199A1 (en) * | 2018-04-03 | 2019-10-03 | Amadeus S.A.S. | Updating cache data |
CN110858204B (zh) * | 2018-08-16 | 2023-09-19 | ***通信集团重庆有限公司 | 查询请求的处理方法、设备和介质 |
FR3086412B1 (fr) | 2018-09-20 | 2020-10-30 | Amadeus Sas | Recalcul des resultats de recherche precalcules |
CN109788305B (zh) * | 2018-12-10 | 2021-03-02 | 北京爱奇艺科技有限公司 | 一种缓存数据的刷新方法及装置 |
CN110795457B (zh) * | 2019-09-24 | 2023-05-05 | 苏宁云计算有限公司 | 数据缓存处理方法、装置、计算机设备和存储介质 |
US11593356B2 (en) * | 2020-09-11 | 2023-02-28 | ForgeRock, Inc. | Never stale caching of effective properties |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6128701A (en) * | 1997-10-28 | 2000-10-03 | Cache Flow, Inc. | Adaptive and predictive cache refresh policy |
US6725333B1 (en) * | 1999-04-22 | 2004-04-20 | International Business Machines Corporation | System and method for managing cachable entities |
AU3638401A (en) | 1999-11-01 | 2001-05-14 | Ita Software, Inc. | Method and apparatus for providing availability of airline seats |
EP1311957A2 (en) * | 2000-07-17 | 2003-05-21 | Eplication Networks Ltd | BANDWIDTH SAVINGS AND QoS IMPROVEMENT FOR WWW SITES BY CATCHING STATIC AND DYNAMIC CONTENT ON A DISTRIBUTED NETWORK OF CACHES |
US7085747B2 (en) * | 2001-09-26 | 2006-08-01 | J Koninklijke Philips Electronics, Nv. | Real-time event recommender for media programming using “Fuzzy-Now” and “Personal Scheduler” |
WO2006046486A1 (ja) * | 2004-10-27 | 2006-05-04 | Nec Corporation | 資源管理システム、資源情報提供方法、及び、プログラム |
JP5387457B2 (ja) * | 2010-03-10 | 2014-01-15 | 富士電機株式会社 | 遠隔監視装置および該装置におけるデータアクセス方法 |
-
2013
- 2013-08-09 IN IN10461DEN2014 patent/IN2014DN10461A/en unknown
- 2013-08-09 KR KR1020157004168A patent/KR101972199B1/ko active IP Right Grant
- 2013-08-09 EP EP13750848.7A patent/EP2885725B1/en active Active
- 2013-08-09 AU AU2013304366A patent/AU2013304366B2/en active Active
- 2013-08-09 ES ES13750848T patent/ES2714676T3/es active Active
- 2013-08-09 WO PCT/EP2013/002390 patent/WO2014026753A1/en active Application Filing
- 2013-08-09 JP JP2015526899A patent/JP6162240B2/ja active Active
- 2013-08-09 CA CA2875735A patent/CA2875735C/en active Active
- 2013-08-09 CN CN201380037889.4A patent/CN104471573B/zh active Active
- 2013-08-09 BR BR112014031390A patent/BR112014031390A2/pt not_active IP Right Cessation
- 2013-08-09 SG SG11201408814XA patent/SG11201408814XA/en unknown
Non-Patent Citations (1)
Title |
---|
None * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11748348B2 (en) | 2019-02-14 | 2023-09-05 | Amadeus S.A.S. | Processing complex database querys |
US11829345B2 (en) | 2019-03-04 | 2023-11-28 | Amadeus S.A.S. | Distributed data processing |
US11947553B2 (en) | 2019-03-04 | 2024-04-02 | Amadeus S.A.S. | Distributed data processing |
WO2022035481A1 (en) * | 2020-08-10 | 2022-02-17 | Google Llc | Anticipatory pre-execution of data queries |
US11341137B1 (en) | 2020-12-04 | 2022-05-24 | Amadeus S.A.S. | Processing search requests |
EP4009188A1 (en) | 2020-12-04 | 2022-06-08 | Amadeus S.A.S. | Processing search requests |
WO2022117787A1 (en) | 2020-12-04 | 2022-06-09 | Amadeus S.A.S. | Processing search requests |
Also Published As
Publication number | Publication date |
---|---|
EP2885725A1 (en) | 2015-06-24 |
BR112014031390A2 (pt) | 2017-06-27 |
CN104471573B (zh) | 2017-07-18 |
KR20150043338A (ko) | 2015-04-22 |
WO2014026753A1 (en) | 2014-02-20 |
AU2013304366A1 (en) | 2015-02-26 |
JP2015531129A (ja) | 2015-10-29 |
IN2014DN10461A (pt) | 2015-08-21 |
SG11201408814XA (en) | 2015-01-29 |
CA2875735C (en) | 2020-09-15 |
JP6162240B2 (ja) | 2017-07-12 |
KR101972199B1 (ko) | 2019-04-24 |
CN104471573A (zh) | 2015-03-25 |
AU2013304366B2 (en) | 2016-05-12 |
CA2875735A1 (en) | 2014-02-20 |
ES2714676T3 (es) | 2019-05-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP2885725B1 (en) | Updating cached database query results | |
US9235620B2 (en) | Updating cached database query results | |
US20160171008A1 (en) | Updating cached database query results | |
EP2842085B1 (en) | Database system using batch-oriented computation | |
US20130073586A1 (en) | Database system using batch-oriented computation | |
EP2911070B1 (en) | Long-term validity of pre-computed request results | |
US20060149713A1 (en) | System, method, and computer program product for improving accuracy of cache-based searches | |
EP3128441B1 (en) | Handling data requests | |
Bosnes et al. | Predicting blood donor arrival | |
US10956955B2 (en) | Managing pre-computed search results | |
EP2698729B1 (en) | Updating cached database query results | |
Yu et al. | Robust team orienteering problem with decreasing profits | |
WO2015124275A1 (en) | Long-term validity of pre-computed request results | |
EP3016000B1 (en) | Managing pre-computed search results | |
US8688488B2 (en) | Method and apparatus for the prediction of order turnaround time in an information verification system | |
CN107004026B (zh) | 管理预先计算的搜索结果 | |
CN112632392A (zh) | 确定候选下车地点的方法、装置、设备、介质和程序产品 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 20141202 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
RIN1 | Information on inventor provided before grant (corrected) |
Inventor name: DANIELLO, RUDY Inventor name: MAILLOT, NICOLAS Inventor name: ISNARDY, LUC Inventor name: CIABRINI, DAMIEN Inventor name: ROBELIN, CHARLES-ANTOINE Inventor name: JANIN, BENOIT Inventor name: LEGRAND, GUILLAUME |
|
DAX | Request for extension of the european patent (deleted) | ||
17Q | First examination report despatched |
Effective date: 20160908 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: GRANT OF PATENT IS INTENDED |
|
INTG | Intention to grant announced |
Effective date: 20180622 |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE PATENT HAS BEEN GRANTED |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R079 Ref document number: 602013048119 Country of ref document: DE Free format text: PREVIOUS MAIN CLASS: G06F0017300000 Ipc: G06F0016000000 |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: FG4D |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: EP |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: REF Ref document number: 1076917 Country of ref document: AT Kind code of ref document: T Effective date: 20181215 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R096 Ref document number: 602013048119 Country of ref document: DE |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: FG4D |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: MP Effective date: 20181212 |
|
REG | Reference to a national code |
Ref country code: LT Ref legal event code: MG4D |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: HR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: NO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20190312 Ref country code: BG Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20190312 Ref country code: LT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: LV Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: FI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: MK05 Ref document number: 1076917 Country of ref document: AT Kind code of ref document: T Effective date: 20181212 |
|
REG | Reference to a national code |
Ref country code: ES Ref legal event code: FG2A Ref document number: 2714676 Country of ref document: ES Kind code of ref document: T3 Effective date: 20190529 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: GR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20190313 Ref country code: RS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: SE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: AL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: NL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: PL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: PT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20190412 Ref country code: IT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: CZ Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: RO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: IS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20190412 Ref country code: SM Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: EE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: SK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R097 Ref document number: 602013048119 Country of ref document: DE |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: DK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: AT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
26N | No opposition filed |
Effective date: 20190913 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: TR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CH Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20190831 Ref country code: MC Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: LI Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20190831 Ref country code: LU Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20190809 |
|
REG | Reference to a national code |
Ref country code: BE Ref legal event code: MM Effective date: 20190831 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20190809 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: BE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20190831 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CY Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 Ref country code: HU Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT; INVALID AB INITIO Effective date: 20130809 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20181212 |
|
P01 | Opt-out of the competence of the unified patent court (upc) registered |
Effective date: 20230524 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20230822 Year of fee payment: 11 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: FR Payment date: 20230824 Year of fee payment: 11 Ref country code: DE Payment date: 20230821 Year of fee payment: 11 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: ES Payment date: 20231027 Year of fee payment: 11 |